VHF profiler observations of winds and waves

in the troposphere during the Darwin Area

Wave Experiment (DAWEX)

R. A. Vincent, A. MacKinnon, and I. M. ReidDepartment of Physics, University of Adelaide, Adelaide, South Australia

M. J. AlexanderColorado Research Associates, Boulder, Colorado, USA

Received 1 March 2004; revised 26 May 2004; accepted 2 June 2004; published 11 August 2004.

[1] AVHF atmospheric radar (wind profiler) was used to study tropospheric winds duringthe Darwin Area Wave Experiment (DAWEX). The profiler, which operated at afrequency of 54.1 MHz, was located at Pirlangimpi (Garden Point) (11.4�S, 130.5�E) onthe Tiwi Islands. Observations were made regularly up to heights near 8 km, withmaximum heights occurring when convective activity was strongest. Mean windsobserved between October and December 2001 are in good agreement with conditions thatprevailed across northern Australia during this period. During the first two intensiveobservation periods (IOP) during October and November, the zonal and meridional windcomponents were westward and northward, respectively, with stronger values inNovember. By the time of IOP3 in mid-December, the zonal flow was eastward, a patternthat is typical of the Australian monsoon. Fluctuations in the three wind components forperiods less than 3 hours are analyzed for IOP2 in November, when strong convectivestorms (‘‘Hectors’’) occurred on all afternoons over the Tiwi Islands. The fluctuations,which are ascribed to convectively generated gravity waves, show a correspondinglystrong diurnal cycle, with horizontal wind variances peaking between 8 and 12 m2s�2 inthe early afternoon in the lower troposphere. Variances are only �2 m2s�2 in the earlymorning hours. A power spectral analysis shows that oscillations with ground-basedperiods between 8 and 17 min are especially prominent during Hector events. The profilerobservations are compared with a numerical model study of gravity wave generation byconvection on 17 November 2001. There is a satisfactory degree of agreement between thebehavior of the model and profiler oscillations, both as a function of height andtime. INDEX TERMS: 3384 Meteorology and Atmospheric Dynamics: Waves and tides; 3314

Meteorology and Atmospheric Dynamics: Convective processes; 3307 Meteorology and Atmospheric

Dynamics: Boundary layer processes; 3374 Meteorology and Atmospheric Dynamics: Tropical meteorology;

KEYWORDS: gravity waves, convection, VHF radar

Citation: Vincent, R. A., A. MacKinnon, I. M. Reid, and M. J. Alexander (2004), VHF profiler observations of winds and waves in

the troposphere during the Darwin Area Wave Experiment (DAWEX), J. Geophys. Res., 109, D20S02, doi:10.1029/2004JD004714.

1. Introduction

[2] The Darwin Area Wave Experiment (DAWEX) wasdesigned to study the generation of gravity waves byconvection and their subsequent propagation into themiddle atmosphere and ionosphere. The focus ofDAWEX was on the so-called ‘‘Hector’’ phenomena,which are intense convective events that occur on adiurnal basis during the premonsoon season over theTiwi Islands north of Darwin, Australia. Three intensiveobservational campaigns (IOPs) were conducted betweenOctober and December 2001 in order to investigate

gravity wave generation by Hector thunderstorms andother convection in the vicinity of Darwin. For moreinformation about the motivation for DAWEX and theinstruments involved, the reader is referred to Hamiltonand Vincent [2000] and K. Hamilton et al. (The DAWEXfield campaign to study gravity wave generation andpropagation, submitted to Journal of GeophysicalResearch, 2004, hereinafter referred to as Hamilton etal., submitted manuscript, 2004).[3] Hector storms occur during the monsoon break flow

characterized by deep subtropical easterly (westward) flowwith a moderate maximum velocity near 3 km height[Keenan et al., 2000]. The deep convective storms areforced along sea breeze fronts, and their orientation isdetermined by the low-level shear.

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 109, D20S02, doi:10.1029/2004JD004714, 2004

Copyright 2004 by the American Geophysical Union.0148-0227/04/2004JD004714$09.00

D20S02 1 of 11

[4] Dynamics also play a role in the spectrum of gravitywaves generated by convection. At least three mechanisms,including mechanical oscillations, latent heat release and theflow over the top of the storm, are involved [see, e.g., Bereset al., 2002, and references therein]. In their study, Beres etal. [2002] note the importance of tropospheric winds incontrolling the strength and direction of waves generated byconvection. Wind shears and associated wave refraction andcritical-level interactions in the troposphere can alter themomentum flux entering the stratosphere.[5] In order to study in detail the dynamics of the

troposphere during DAWEX, a small VHF radar windprofiler was established at Pirlangimpi (Garden Point) onthe northwestern part of Melville Island, adjacent to Apsleystrait, the narrow body of water that separates Melville fromBathurst Island. A map showing the location of the radarand the relevant coordinates are given in the accompanyingoverview paper by Hamilton et al. (submitted manuscript,2004). The radar allowed the basic flow in the lower andmiddle troposphere to be characterized during the IOPs, andthe excellent height and time resolution enabled the wavefield to be investigated on a diurnal basis.[6] The paper is organized as follows. Section 2 describes

the radar system and its calibration. Wind measurements arediscussed in section 3, with descriptions of the mean windsobserved during all IOPs and wave motions in IOP2. Theresults are discussed in section 4 with a comparison betweenprofiler wave observations on 17 November and numericalmodel results.

2. Equipment and Calibration

[7] The VHF radar used in the DAWEX campaignswas similar to the small system described in the work of

Vincent et al. [1998]. This was designed to study tropo-spheric dynamics down to heights near 300 m in theboundary layer. However, the DAWEX radar was morepowerful and used larger antennas for transmission andreception than the original system so that it had a betterheight coverage. The coverage was also improved byoperating in a humid tropical environment since verticalgradients in humidity play an important role in determin-ing radar reflectivities.[8] The radar array was used for both transmission and

reception. As shown in Figure 1, it consisted of threesubgroups, each composed of nine three-element Yagiantennas arranged in a 3 � 3 matrix. Each Yagi wasoriented at 45� to the major axes of the array so that thebasic spacing was 0.5 wavelength. This ensured a compactarrangement and minimized sidelobes and associatedground clutter, which is important when receiving echoesfrom low altitudes. All antennas were phased to pointvertically, with transmission on all 27 antennas in order toprovide as narrow a beam as possible. The half-power-full-beamwidth (HPFB) of each subgroup was 32�, while theHPFB of the transmit antenna was 18�. Each subgroup wasconnected to its own receiver, and horizontal winds weremeasured using the spaced-antenna method [Briggs, 1984],while vertical velocities were computed from the Dopplershifts of the received signals.[9] During DAWEX the radar operated in two modes. A

low-level mode that used a 750 ns pulse to obtain 100 mheight resolution over a height range between 300 m and3.7 km and a high-level mode that used a 4 ms length pulseto measure winds starting at 2 km with a height resolutionof 600 m with data oversampled every 300 m. Table 1summarizes the operating parameters used during theDAWEX campaigns.[10] The power received by a VHF radar depends on the

type of scattering or reflecting processes involved, whichrange from volume scatter to specular reflection. However,the common feature is a dependence on the gradient of radiorefractive index, M [Doviak and Zrnic, 1993]. We made useof this dependence to check the radar height calibration.Correct range determination requires knowledge of signaldelays through the radar system, including the cables con-necting the antennas, which are not always easy to estimateaccurately. As part of the calibration process, values of M2

were derived from simultaneous high-resolution radiosondeobservations made every 3 hours from Garden Point duringthe IOPs (see T. Tsuda et al., Characteristics of gravitywaves with short vertical wavelengths observed with radio-sonde and GPS occultation during DAWEX (Darwin Area

Figure 1. Plan view of antenna array. The groups labeledA, B, and C denote the antenna subgroups used forreception. The whole array was used for transmission.

Table 1. VHF Profiler Operating Parameters

Operating Parameters Low Mode High Mode

PRF, Hz 20,000 8000Pulse length, m 100 600Range, km 0.3–3.8 2.0–10.0Range sampling, m 100 300Receiver bandwidth, kHz 404 253Coherent integrations 1000 400Number of samples 1100 1100Acquisition length, s 55 55Nyquist velocity, m s�1 ±27.7 ±27.7

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

2 of 11

D20S02

Wave Experiment), submitted to Journal of GeophysicalResearch, 2004, for details).[11] The refractive index gradient is given by

M ¼ @n

@z

¼ �77� 10�6 p

T

@ ln q@z

� �

� 1þ 15500q

T1� 1

2

@ ln q=@z

@ ln q=@z

� �� �; ð1Þ

where n is the refractive index. Values ofM2 were computedfrom radiosonde profiles of temperature T, pressure p

(measured in hPa), potential temperature q, and specifichumidity q. The soundings had a 2-s time resolution,equivalent to an approximate 10 m height resolution.Comparisons of vertical profiles of echo strength and M2

showed that the radar ranges were about 100 m too high(Figure 2). Accordingly, all measurements recorded by theradar have been shifted down by this amount. All heightsmentioned from now on are corrected heights.[12] After the radar ranges were corrected, the profiler

winds were compared with the values determined from theradiosonde flights. Comparisons made over all campaignsare shown in Figure 2. There is good agreement between thetwo sets of observations, especially in direction. However,

Figure 2. (top) Examples of comparisons of M2 computed from radiosonde observations and radarpower profiles (dotted) before and (solid) after moving downward by 100 m. (bottom) Comparisons of(left) speed and (right) direction of tropospheric winds measured by the VHF profiler and radiosondeobservations at Garden Point during DAWEX.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

3 of 11

D20S02

in common with previous wind comparisons betweenspaced antenna radars and radiosondes, the speeds derivedfrom the profiler underestimate the actual values by up to9%.[13] The system alternated between low and high mode

every 1 min, so the nominal time resolution for each modewas 2 min. However, the actual time resolution achieveddepended on the presence of suitable scattering irregulari-ties. Figure 3 shows the percentage acceptance rate ofhorizontal wind measurements during the three IOP cam-paigns. The acceptance rate varied between 80 and 95% upto heights near 4 km and steadily decreased above thisheight, with the 50% acceptance rate occurring near 6 km.

As discussed later, there is also a diurnal variation in echostrengths and height coverage.

3. Observations

3.1. Mean Winds

[14] The profiler observations show that the mean windschanged systematically between the first campaign inOctober and the final campaign in December, as illustratedin Figure 4. In October the zonal wind averaged over thewhole campaign is westward (easterly) at all heights, with apeak value of 6 m s�1 near 2 km. The meridional compo-nent is northward (southerly) above 2 km and weaklysouthward below this height.[15] In November the westward zonal winds had

increased in strength, peaking near 3 km altitude at12 m s�1, although at heights below 1 km the winds hadreversed to weak eastward (westerly) winds. The meridionalcomponent was northward at all heights up to 8 km, withmaximum values of 5 m s�1. These conditions are typical ofmonsoon break period when Hectors are formed [Keenan etal., 2000].[16] By the time of the December IOP the zonal winds

had reversed direction, becoming eastward (westerly) at allheights of observation with mean values of about 5 m s�1.This kind of flow is typical of the monsoon over northernAustralia. The meridional winds show a weak shear, chang-ing from a 1–2 m s�1 near the ground to a weak southwardflow of similar magnitude near 8 km.[17] Examination of the mean winds during each IOP

showed little evidence for any systematic variations withtime. For reference, Hamilton et al. (submitted manu-script, 2004) provide a time-height plot of hourly averagezonal winds during IOP2 derived from the profiler

Figure 3. Percentage acceptance rate of horizontal windsderived from VHF profiler during the three DAWEX IOPcampaigns.

Figure 4. Vertical wind profiles of (dotted) mean zonal, (dashed) meridional, and (solid) vertical windcomponents in the lower troposphere during the intensive observation campaigns.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

4 of 11

D20S02

observations. This shows that there is little significantdiurnal variation of the winds. In summary, the meanhorizontal winds observed by the profiler during eachIOP were in good agreement with the general flowpatterns observed over northern Australia, as discussedin the work of Hamilton et al. (submitted manuscript,2004).[18] Vertical velocities were measured from the Doppler

shifts of the backscattered echoes. At all times the meanvertical velocities were very small and, on average, zero.Examination of the raw data did indicate short periods of

significant updrafts, with peak velocities of a few meters persecond during strong convection.

3.2. Wave Observations

[19] As noted above, the effective wind sampling intervalis greater than 2 min at each height owing to missing data.To provide time series of the winds, each wind componentwas averaged over 4 min intervals, and missing data werelinearly interpolated over. In order to study variations thatmight be due to gravity waves generated by Hector andother convective storms in the vicinity of Darwin, the data

Figure 5. Stacked plots of (top) zonal, (middle) meridional, and (bottom) vertical wind componentsduring the November campaign. The height resolution is 100 m up to 3.6 km and 300 m above thatheight. The time series are filtered between periods of 12 and 180 min.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

5 of 11

D20S02

at each height were filtered with a bandpass between 12 and180 min. Time series of the filtered wind components areplotted in Figure 5 as a function of height. The data areplotted with a height spacing of 100 m up to 3.6 km, andthen at a spacing of 300 m.[20] Several features are evident in Figure 5. First, there is

a strong diurnal modulation of height coverage. Winds areoften measured up to heights near 8 km during periodsbetween �0600 and 1800 UT (1530–0330 LT). Thischange is due to increased reflectivities brought about bythe uplift of water vapor into the middle troposphere by theafternoon convective storms. As shown by equation (1), therefractivity is a strong function of specific humidity q.Another factor in improving the height coverage is theincreased levels of turbulence associated with the convec-tion and the corresponding increase in strength of thescattering irregularities.[21] The second feature is the absence of wind mea-

surements at the lower heights during times when con-vection is strong. For example, for a period centered on0600 UT (1530 LT) on 19 November, there is an absenceof wind measurements up to heights near 2 km. There aretwo reasons for these gaps. Any convective cells passingover the radar produced strong small-scale variations inwinds, causing spectral broadening of the echoes. Corre-spondingly, the correlation functions used in the spacedantenna full correlation analysis (FCA) are narrow, andthe analysis often breaks down. The second reason for alack of wind measurements was caused by the radar’slocation adjacent to Apsley strait. Radar backscatter fromsea waves usually produces narrow spectral lines that arerelatively easy to remove in the spectral domain before

the FCA is carried out. However, during convectivestorms, the strong winds produce enhanced sea surfaceroughness, which led to strong spectral broadening of thesea clutter. It was often impossible to distinguish theclutter part of the spectrum from the atmospheric com-ponent, which itself was broadened owing to the strongerturbulent motions. Under these conditions the wind anal-ysis often broke down or produced spurious results.

Figure 6. Frequency spectra for the (dark solid line) zonal, (dashed line) meridional, and (light solidline) vertical perturbation motions observed in the 2.3 and 2.7 km height region. The spectra are averagedover all days of observation in IOP2 for the time intervals shown. For reference, the straight line in eachpanel has a slope of f�5/3.

Figure 7. Mean square amplitudes of the (dark solid line)zonal, (dashed line) meridional, and (light solid line)vertical wind components in the 8-min to 3-hour periodband observed between 2 and 2.5 km during IOP2.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

6 of 11

D20S02

[22] Notwithstanding the complications in making windmeasurements when strong convection was present, it ispossible to determine gravity wave amplitudes. First, powerspectral analyses were carried out for each wind componentto study the distribution of wave energy as a function offrequency. In order to study the evolution of the spectra as afunction of time, the spectra were computed in 3-hour timeintervals, each overlapped by 1.5 hours. The number ofdegrees of freedom, and hence spectral reliability, wasincreased by averaging spectra together over a range ofheights and for all days of observation. Figure 6 showsspectra computed for the 0300–0600 UT (1230–1530 LT)

and 1500–1800 UT (0030–0330 LT) intervals in the 2.3–2.7 height range. Each spectral estimate has a notional50 degrees of freedom associated with it, although the useof 100 m radar pulses means that the observations made atadjacent range gates may not be fully independent.[23] There is a clear difference in both the spectral shape

and amplitudes between motion fields observed in the twointervals that are 12 hours apart. First, in the early afternooninterval, the spectral energy of all three wind components is1–2 orders of magnitude larger than the values observed inthe early morning interval. Second, the horizontal motions,and particularly the zonal (u0) component, have a broad

Figure 8. Height profiles of rms amplitudes for motions in the 12–180 min period range for theintervals (top) between 0300 and 1100 UT (between 1230 and 2030 LT) and (middle) between 1500 and2300 UT (between 0030 and 0830 LT). (bottom) Gravity wave amplitudes derived from a numericalsimulation for 0300–1100 UT, 17 November 2004 [Alexander et al., 2004].

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

7 of 11

D20S02

peak at frequencies between 2 � 10�3 and 1 � 10�3 Hz,i.e., at ground-based periods between �8 and 17 min. Thispeak is even more evident when the spectra are plotted inarea preserving form.[24] It should be pointed out that differential vertical

motions across the radar beam can appear as spurioushorizontal motions in the spaced antenna analysis [Briggs,1980] and can distort a spectrum of gravity wave motions[Rastogi et al., 1996]. The effect is significant when thedifferential vertical motions have scales of the diameter ofthe of the area subtended by the radar beam, which at aheight of 2–3 km is �1–1.5 km. It seems unlikely thatgravity waves would have horizontal wavelengths as shortas this.[25] The area under each spectrum gives the mean square

amplitude of the motions. Time series for IOP2 of the meansquare amplitude of each wind component observed in theheight range 2–2.5 km are provided in Figure 7. Results areplotted for 3-hour time intervals, overlapped by 1.5 hours.The figure brings out more clearly the strong diurnal cyclein activity evident in Figure 5. Peak rms values rangebetween about 2 and 2.5 m s�1 for both u0 and v0 andbetween about 0.5 and 1 m s�1 for w0.

4. Discussion

[26] The results presented above need to be placed incontext with respect to the meteorological conditions thatpertained at the time in IOP2. Using C-pol radar reflec-

tivities as a guide, Hector-like storms appeared on theTiwi Islands every afternoon between 15 and 19 Novem-ber (Hamilton et al., submitted manuscript, 2004). Typi-cally, they had maximum development between 0330 and0600 UT (between 1300 and 1530 LT). In general, thestorm cells passed either to the south or north of GardenPoint by distances of up to 20–40 km (see, e.g.,Hamilton et al., submitted manuscript, 2004, Figure 10).However, on 16 November a strong convective stormpassed almost over Garden Point at about 0430 UT(1400 LT) and again on 17 November at about 0410 UT(1340 LT). In evening hours, squall lines appeared overcontinental Australia, usually being quite intense at about1200 UT (2130 LT) and traveled northwestward towardthe Tiwi Islands.[27] On the basis of the C-Pol reflectivities the Hector

on the 18 November was the strongest storm of IOP2(P. T. May, private Communication, 2004), which corre-lates with the strongest peak in variances (Figure 7). It isnot easy to assess how much of the wind fluctuationsmeasured by the VHF profiler can be ascribed to con-vective motions and how much due to waves. Theproximity of the storm to the radar on the 18 Novembermeans that the motions could be due to a combination ofconvective motions and wave motions as the stormpassed just to the north of the radar on that date.However, the fact that storms passed some 20 km ormore from the profiler on other days suggest that wavemotions are responsible for peak variances of 12 m2s�2 inhorizontal motions and 0.5–1 m2s�2 in the verticalmotions.[28] The frequency spectra shown in Figure 6 suggest

that much of the wave energy is concentrated at ground-based periods between 8 and 17 min, which is close tothe buoyancy frequency of 9–10 min in the lowertroposphere during IOP2. Various numerical modelingstudies that use realistic environments for the initiationand development of Hectors show dominant gravitywaves with wavelengths of 15–25 km and intrinsicperiods between 15 and 20 min [Piani et al., 2000; Laneet al., 2001; Lane and Reeder, 2001]. The rather mono-chromatic wave fields that appeared above the tropopausein these models were ascribed by Lane and Reeder[2001] to convective overshoot of air parcels and oscil-lations around the level of neutral buoyancy.[29] Alexander et al. [2004] in an accompanying paper

describe a numerical simulation of gravity wave genera-tion by convection during DAWEX. They used C-Polweather radar reflectivities (see Hamilton et al., submittedmanuscript, 2004 for details) to delineate the temporaland geographic variability of gravity wave forcing bylatent heat release during convection. The model, whichis uses a 400 � 400 km domain centered on the C-Polradar, has a 2-km resolution in the horizontal and 0.25 kmin the vertical. Alexander et al. [2004] focus on a 7-hourperiod on 17 November 2001 from 0300 to 0950 UT(1230 to 1920 LT). The conditions on this day are typicalof all days during IOP2, i.e., there were Hectors in theafternoon and continental squall lines in the evening. Inthe area covered by the C-Pol radar the maximumvolumetric heating rates peaked at about 0430 and0830 UT.

Figure 9. Comparison for a height of 2.5 km of (top)observed and (bottom) model wave amplitudes for the (darksolid line) zonal, (dashed line) meridional, and (light solidline) vertical wind components. Note that for ease ofcomparison, the vertical velocities have been increased by afactor of 4.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

8 of 11

D20S02

Figure 10. Time-height cross sections of perturbation amplitudes in the 20–180 min range from (left)profiler and (right) model. The top panels show u0, middle panels show v0, and bottom panels show w0.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

9 of 11

D20S02

[30] Figure 8 shows height variations of wave amplitudesderived for 17 November. In order to compare in moredetail with results derived from a numerical model (seebelow), the data were divided into two intervals, onecentered on the most active period between 0300 and1100 UT (between 1230 and 2030 LT) and the othercentered on the early morning hours (1500–2300 UT or2330–0830 LT), when convection was not strong. Duringthe active period in midafternoon (local time), rms ampli-tudes were about 1.5 m s�1 for both horizontal windcomponents, while the value for the vertical componentwas �0.4 m s�1. Both u0 and v0 show some height structure,while w0 is almost constant with height. Amplitudes duringthe early morning hours are 2–3 times smaller for allcomponents than in the afternoon period and the heightprofiles are smoother.[31] The tropospheric perturbations in the model at the

model grid point closest to the Garden Point profiler can becompared to the radar wind values. Model results for the0300–1100 UT time interval are shown in the bottom panelof Figure 8. Amplitudes of the horizontal perturbationmotions lie in the range 0.1–0.16 m s�1, while the verticalmotion amplitudes are about 0.03 m s�1. Interestingly, theheight structure of the model and observed winds are rathersimilar. However, the model values are approximately afactor of 10 smaller than the profiler measurements, an issuethat is discussed further below.[32] Figure 9 compares time series of the observed and

model winds for a height of 2.5 km. The data have beenfiltered to remove periods shorter than 24 min and longerthan 3 hours. Despite the filtering, the observations showdistinct short period oscillations that are not evident in themodel results. This difference is likely associated with shorthorizontal-scale motions in the data that are either advectedor propagate over the radar site. The model resolution isonly 2 km, and the numerical dissipation effectively dampshorizontal-scale fluctuations at scales 10 km and shorter.[33] What is particularly apparent, however, is the simi-

larity between the model and observed fluctuations. Thesesimilarities are especially evident in the u0 components,which show very similar phases, evolution with time, andare dominated by �2-hour period waves. The v0 compo-nents, however, show stronger oscillations at periods near1 hour in the first part of the record, but which are absent inthe last part from about 0830 UT onward. The verticalvelocity components similarly have larger amplitudes inthe interval before 1800 LT. Again, it is noteworthy that theobserved amplitudes are some 5–10 times larger than themodel values.[34] Perturbations as a function of height and time are

displayed in Figure 10. The observations were filtered toinclude fluctuations with periods 20 min to 3 hours, andthe model for fluctuations shorter than 3 hours. Clearly,there are some differences between the observations andmodel results owing to the somewhat different spectrumof oscillations that are present in each data set (as notedabove the model effectively filters out higher frequencymotions). The broader bandwidth of the radar resultsleads to less coherence in the motions with height andtime. Despite the differences, there are some remarkablysimilar features in both the model and the data. First isthe time evolution, with much larger amplitude, vertically

coherent structures that appear only in the first portion ofthe time record, and which are absent in both data andmodel at later times. There is also a very similar shortvertical-scale feature in the second half of the time periodin the zonal wind perturbations that shows downwardphase propagation. This may be a signature of a zonallypropagating wave that appears at both the same time andplace in both the data and the model. The meridional andvertical motions tend to show more vertical structure,although there is evidence of vertical phase tilt in theobserved vertical component at about 0630 UT.[35] Despite the similarities the wave amplitudes in the

model are roughly 10 times smaller than the observations.At face value this difference would suggest that the modelinput heating and output wave amplitudes should be scaledupwards by about a factor of 10. The problem is that theexact conversion between the latent heating and wavegeneration is unknown.

5. Summary and Conclusions

[36] The VHF profiler observations presented hereallowed us to explore gravity wave variability in thelower troposphere in the vicinity of intense deep convec-tive storms (Hectors) over the Tiwi Islands in northernAustralia. During the November 2001 campaign forDAWEX a strong diurnal cycle in wave activity wasobserved, with peak wave amplitudes reached in mid-afternoon. For waves with periods in the 8–180 min periodrange, an approximately 10:1 ratio between the largest andsmallest variances is found. Oscillations with ground-basedperiods between 8 and 17 min are especially prominent inthe early afternoon measurements. In general, during theNovember campaign, areas of strongest convection passedeither to the north or south by distances ranging from 10 to30 km, so these observations represent the ‘‘near-field’’response to the convection.[37] A case study for 17 November 2001 compares

radar observations with results from a numerical modelusing weather radar reflectivities to help simulate gravitywave forcing due to latent heat release in convection[Alexander et al., 2004]. There are significant similaritiesbetween the model and observed fluctuations, whichprovide encouragement to carry out further comparisons.An important issue requiring resolution is the ‘‘calibra-tion’’ of the model wave amplitudes. One approach thisissue is to exploit the ability of VHF boundary layerprofilers to measure raindrop size distributions with goodtime and height resolution down to heights near thesurface [Lucas et al., 2004]. Using a VHF profiler inconjunction with cloud and weather radars, it would bepossible to investigate convective cloud microphysics inmore detail. This would provide better understanding ofthe relationship between weather radar reflectivities andlatent heat release, in turn providing better specificationsof latent heat release for model simulations of gravitywaves.

[38] Acknowledgments. We gratefully acknowledge the TiwiLand Council for their support and permission to operate the profiler atPirlangimpi. Helpful discussions with P. T. May are also appreciated. Thiswork was supported by Australian Research Council grants A69802414 andX00001692.

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

10 of 11

D20S02

ReferencesAlexander, M. J., P. T. May, and J. H. Beres (2004), Gravity waves generatedby convection in the Darwin area during the Darwin Area Wave Experi-ment, J. Geophys. Res., 109, D20S04, doi:10.1029/2004JD004729, inpress.

Beres, J. H., M. J. Alexander, and J. R. Holton (2002), Effects of tropo-spheric wind shear on the spectrum of convectively generated gravitywaves, J. Atmos. Sci., 59, 1805–1824.

Briggs, B. H. (1980), Radar observations of atmospheric winds andturbulence: A comparison of techniques, J. Atmos. Terr. Phys., 42,823–833.

Briggs, B. H. (1984), The analysis of spaced sensor records by correlationtechniques, in Handbook for MAP, 13, edited by R. A. Vincent, pp. 166–186, SCOSTEP Secretariat, Urbana, Ill.

Doviak, R. J., and D. S. Zrnic (1993), Doppler Radar and WeatherObservations, 2nd ed., Academic, San Diego, Calif.

Hamilton, K., and R. A. Vincent (2000), Experiment will examine gravitywaves in the middle atmosphere, Eos Trans. AGU, 81, 517.

Keenan, T., et al. (2000), The Maritime Continent Thunderstorm Experi-ment (MCTEX): Overview and some results, Bull. Am. Meteorol. Soc.,81, 2433–2455.

Lane, T. P., and M. J. Reeder (2001), Convectively generated gravitywaves and their effect on the cloud environment, J. Atmos. Sci., 58,2427–2440.

Lane, T. P., M. J. Reeder, and T. L. Clarke (2001), Numerical modelling ofgravity wave generation by deep tropical convection, J. Atmos. Sci., 58,1249–1274.

Lucas, C., A. D. MacKinnon, R. A. Vincent, and P. T. May (2004), Rain-drop size distribution retrievals from a VHF boundary layer profiler,J. Atmos. Oceanic Technol., 21, 45–60.

Piani, C., D. Durran, M. J. Alexander, and J. R. Holton (2000), A numericalstudy of three-dimensional gravity waves triggered by deep tropical con-vection and their role in the dynamics of the QBO, J. Atmos. Sci., 57,3689–3702.

Rastogi, P. K., E. Kudeki, and F. Surucu (1996), Distortion of gravity wavespectra of horizontal winds measured in atmospheric radar experiments,Radio Sci., 31, 105–118.

Vincent, R. A., S. Dullaway, A. MacKinnon, I. M. Reid, F. Zink, P. T. May,and B. H. Johnson (1998), A VHF boundary layer radar: First results,Radio Sci., 33, 845–860.

�����������������������M. J. Alexander, Colorado Research Associates, 3380 Mitchell lane,

Boulder, CO 80301, USA. (alexand@cora.nwra.com)A. MacKinnon, I. M. Reid, and R. A. Vincent, Department of Physics,

School of Chemistry and Physics, University of Adelaide, Adelaide 5005,South Australia. (andrew.mackinnon@adelaide.edu.au; iain.reid@adelaide.edu.au; robert.vincent@adelaide.edu.au)

D20S02 VINCENT ET AL.: VHF PROFILER OBSERVATIONS IN DAWEX

11 of 11

D20S02